Resource utilization by server systems is dynamic, as computational demand on the server systems varies. Spikes in computational demand on a server system can lead to resource overloading, resulting in system shutdown, failure, and/or data loss. For example, an uninterruptable power supply may shutdown responsive to overloading causing thousands of servers to lose power. In order to prevent resource overloading, server systems are often run significantly under-capacity in anticipation of worst case computational demand. As a result, server systems are inefficient because the capacity that is reserved to compensate for computational demand spikes is unavailable for normal server activity.